The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection through its RAG pipeline templates.
Ingestion points: External data from Tavily and Exa searches is stored in variables (e.g., SEARCH_RESULT, CONTENT, EVIDENCE) and then passed into model prompts within SKILL.md examples.
Boundary markers: The prompt templates use simple text headers (e.g., 'Search Results:', 'Source 1 (Tavily):') which can be easily bypassed if the retrieved web content contains instructions designed to override the agent's behavior.
Capability inventory: The skill possesses the capability to execute any platform command via infsh as specified in the allowed-tools field.
Sanitization: There is no evidence of sanitization, escaping, or filtering of the external content before it is interpolated into the prompts.
[COMMAND_EXECUTION]: The provided bash examples use unsafe shell variable interpolation.
The examples construct JSON payloads for the infsh CLI using double-quoted strings (e.g., "{\"prompt\": \"... $SEARCH\"}"). If the retrieved search content contains double quotes, backslashes, or backticks, it will cause the JSON to be malformed or potentially lead to unexpected shell behavior during command execution.
[EXTERNAL_DOWNLOADS]: The skill correctly references its own ecosystem tools.
It suggests installing the inference-sh/skills@agent-tools package via npx, which is consistent with the skill's stated purpose and authored environment.

ai-rag-pipeline